Operand conversion optimization

• US 7,051,060 B2
• Filed: 09/28/2001
• Issued: 05/23/2006
• Est. Priority Date: 09/28/2001
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Ford, Richard L.
• Primary Examiner(s): MALZAHN, DAVID H
• Application Number:
  US09/966,701
• Publication Number:
  US 20030065698A1
• Time in Patent Office: 1,698 Days
• Field of Search: 708/490, 708/495
• US Class Current: 708/495
• CPC Class Codes: G06F 7/483 Computations with numbers represented by a non-linear combination of denominational numbers, e.g. rational numbers, logarithmic number system or floating-point numbers G06F7/4806, G06F7/4824, G06F7/49, G06F7/491, G06F7/544 take precedence
• CPC Class Codes: G06F 8/443 Optimisation

Abstract

According to the invention, optimization of an application by elimination of redundant operand conversions is disclosed. According to one embodiment, the optimization comprises receiving an application that includes one or more operations, with one or more operands of the operations being converted from a first format to a second format before performing an operation; determining the origin of the one or more operands that are converted from the first format to the second format; and if the origin of any of the one or more operands that are converted from the first format to a second format is a conversion from the second format to the first format, then eliminating the redundant conversion from the second format to the first format and from the first format to the second format.

Claims

1. A method comprising:

expanding an application into a series of operations, the series of operations including a first operation and a second operation;

examining the series of operations to determine that;

a result of the first operation is an operand of the second operation, and

the series of operations includes a first conversion of the result of the first operation from a first format to a second format after the first operation and a second conversion of the operand of the second operation from the second format to the first format prior to the second operation; and

eliminating the first conversion and the second conversion from the series of operations.

2. The method of claim 1, further comprising partially compressing the series of operations after the elimination of the first conversion and the second conversion.

3. The method of claim 1, further comprising determining that the series of operations includes a third conversion of a result from the second format to the first format that is immediately followed by a fourth conversion of an operand from the first format to the second format in a first case and that is not immediately followed by the fourth conversion in a second case.

4. The method of claim 3, wherein the third conversion and the fourth conversion occur within a loop, and further comprising moving the third conversion and the fourth conversion from inside the loop to outside the loop.

5. The method of claim 1, wherein the first format is a unpacked floating point format and the second format is an packed floating point format.

6. A method comprising:

receiving an application, the application including one or more floating point calculations;

expanding the application, the expansion of the application comprising expanding the floating point calculations into a series of floating point operations on floating point operands, the floating point operations including a conversion of an operand of a floating point operation from a packed format to an unpacked format prior to the performance of the floating point operation and converting the result of a floating point operation from the unpacked format to the packed format after the performance of the floating point operation;

determining that a result of a first floating point operation of the series of floating point operation is an operand of a second floating point operation of the series of floating point operations; and

eliminating the conversion of the result of the first floating operation from the unpacked format to the packed format and the conversion of the operand of the second floating point operation from the packed format to the unpacked format.

7. The method of claim 6, wherein the packed format is a standard floating point format.

8. The method of claim 7, wherein the standard floating point format is in accordance with Standard 754 of the Institute of Electrical and Electronic Engineers.

9. The method of claim 6, wherein the unpacked format is an operand format that may be used in performing floating point emulation of a floating point operation.

10. The method of claim 9, wherein a conversion of a floating point operand from the packed format to the unpacked format comprises a conversion of the floating point operand to one or more integer values.

11. The method of claim 6, further comprising partially compressing the remaining floating point operations, if any, after the elimination of the conversions.

12. The method of claim 6, further comprising determining that the series of floating point operations includes a third conversion of a result from the second format to the first format that is immediately followed by a fourth conversion of an operand from the first format to the second format in a first case and that is not immediately followed by the fourth conversion in a second case.

13. The method of claim 12, wherein the first conversion and the fourth conversion are within a loop, and the sequences of instructions include instructions that, when executed by a processor, cause the processor to perform operations comprising moving one or more operand conversions from inside the loop to outside the loop.

14. A machine-readable medium having stored thereon data representing sequences of instructions that, when executed by a processor, cause the processor to perform operations comprising:

expanding an application into a series of operations, the series of operations including a first operation and a second operation;

examining the series of operations to determine that;

a result of the first operation is an operand of the second operation, and

the series of operations includes a first conversion of the result of the first operation from a first format to a second format after the first operation and a second conversion of the operand of the second operation from the second format to the first format prior to the second operation; and

eliminating the first conversion and the second conversion from the series of operations.

15. The medium of claim 14, wherein the sequences of instructions include instructions that, when executed by a processor, cause the processor to perform operations comprising partially compressing the series of operations after the elimination of the first conversion and the second conversion.

16. The medium of claim 14, wherein the sequences of instructions include instructions that, when executed by a processor, cause the processor to perform operations comprising determining that the series of operations includes a third conversion of a result from the second format to the first format that is immediately followed by a fourth conversion of an operand from the first format to the second format in a first case and that is not immediately followed by the fourth conversion in a second case.

17. The medium of claim 16, wherein the third conversion and the fourth conversion are within a loop, and wherein the sequences of instructions include instructions that, when executed by a processor, cause the processor to perform operations comprising instructions for moving the third and fourth conversions from inside the loop to outside the loop.

18. The medium of claim 14, wherein the first format is a unpacked floating point format and the second format is an packed floating point format.

19. A machine-readable medium having stored thereon data representing sequences of instructions that, when executed by a processor, cause the processor to perform operations comprising:

receiving an application, the application including one or more floating point calculations;

expanding the application, the expansion of the application comprising expanding the floating point calculations into a series of floating point operations on floating point operands, the floating point operations including a conversion of an operand of a floating point operation from a packed format to an unpacked format prior to the performance of the floating point operation and converting the result of a floating point operation from the unpacked format to the packed format after the performance of the floating point operation;

determining that a result of a first floating point operation of the series of floating point operation is an operand of a second floating point operation of the series of floating point operations; and

eliminating the conversion of the result of the first floating operation from the unpacked format to the packed format and the conversion of the operand of the second floating point operation from the packed format to the unpacked format.

20. The medium of claim 19, wherein the packed format is a standard floating point format.

21. The medium of claim 20, wherein the standard floating point format is in accordance with Standard 754 of the Institute of Electrical and Electronic Engineers.

22. The medium of claim 19, wherein the unpacked format is an operand format that may be used in performing floating point emulation of a floating point operation.

23. The medium of claim 22, wherein a conversion of a floating point operand from the packed format to the unpacked format comprises a conversion of the floating point operand to one or more integer values.

24. The medium of claim 19, wherein the sequences of instructions include instructions that, when executed by a processor, cause the processor to perform operations comprising partially compressing the remaining floating point operations, if any, after the elimination of the conversions.

25. The medium of claim 19, wherein the series of instructions further comprises instructions for determining that the series of floating point operations includes a third conversion of a result from the second format to the first format that is immediately followed by a fourth conversion of an operand from the first format to the second format in a first case and that is not immediately followed by the fourth conversion in a second case.

26. The medium of claim 25, wherein the third conversion and the fourth conversion are within a loop, and wherein the sequences of instructions include instructions that, when executed by a processor, cause the processor to perform operations comprising moving the third and fourth conversions from inside the loop to outside the loop.